ISSN (Print) - 0012-9976 | ISSN (Online) - 2349-8846

A+| A| A-

Migration, Employment Status and Poverty

This paper analyses the pattern of migration in urban areas and its socio-economic correlates. The analysis is based on the National Sample Survey's reports of employment and unemployment pertaining to the latest rounds, which provide information on migration. Economic deprivation is not the most critical factor for migration decisions, even for seasonal migrants. People migrate out of both poor and rich households, although the reasons for migration and the nature of jobs sought by them are different. Rural-urban migrants have a greater risk of being below the poverty line than the urban-urban migrants, but both report a lower risk than non-migrants. The probability of a person being poor is low in a large city compared to any other urban centre, irrespective of the migration status, age, number of subsidiary activities undertaken, etc. The results indicate that migration has been a definite instrument of improving economic well-being and escaping from poverty. The probability of being poor is much less among the migrants compared to the local population, in all size classes of urban centres.


Migration, Employment Status and Poverty

An Analysis across Urban Centres


Economic and Political WeeklyJanuary 27, 2007300The medium category cities/towns, with population between50,000 and 1 million, reported poverty levels of 28 per cent in1993-94 which has gone down to 20 per cent in 1999-2000. Thecorresponding figures in small towns, those with 50,000 or lesspeople, are as high as 33 per cent and 24 per cent respectively,slightly higher than even in rural areas. There are, thus, reasonsto be concerned also about the poverty situation in the lowercategories of urban settlements, as much as in rural areas.In the context of the above, the present paper attempts toidentify the key determinants of poverty for individuals andhouseholds residing in different size class of urban centres, usinginformation on consumption expenditure, employment status andother socio-economic and locational characteristics, based on thedata from the National Sample Survey (NSS). In particular, anattempt has been made to estimate the impact of the level ofeducation, migration status, occupation, nature of employmentand city size on the probability of a household/person fallingbelow the poverty line. One would expect a lower incidence ofpoverty in larger cities than in smaller towns as employment andother economic opportunities are more in the former. These citiesprovide better social and physical infrastructure, includingeducationalfacilities, which results in higher factor productivity[Sviekauskas 1975]. Correspondingly, returns to education arelikely to be higher in large cities than in small towns. Persons inregular and self-employment understandably are better placedthan the casual workers. A large majority among the former areexpected to be above the poverty line. It is also argued thatmigration,ingeneral, is linked to expected future social and eco-nomic pay-offs in comparison with those in the present. Hence,a person is likely to have a greater chance of escaping povertyafter migration if it is his/her free choice. The motivation to sendout a migrant would be stronger for a poor than a rich householdalthough the capacity to “afford” migration is higher in case ofthe latter.Given these hypotheses and controversies around these, thepresent paper analyses the incidence of poverty across differentsize class of urban centers, incorporating the nature of themigrant,occupational characteristics, levels of consumptionexpenditure and educational attainments as explanatory param-eters. The SectionII presents the data base and methodology ofanalyses, elaborating the framework and rationale of the statis-tical exercise and highlighting its limitations. Section III probesinto the pattern of migration across various socio-economiccategories, as also the incidence of poverty by the nature ofmigration, through cross tabulation of information. The tableshave been generated using the published data in various NSSrounds, as also the unit level unpublished data for the 55th round.The results of logistic regression analysis attempting to determinethe impact of identified socio-economic factors on the probabilityof being poor, after controlling for the influence of other factors,are presented inSection IV. Section V summarises the conclusionsand discussestheir implications in the contemporary context ofdevelopment.IIExplanatory Framework:Database and MethodologyThe population census provides detailed information on trendsand pattern of migration at the district level, as also for classI cities. It is, however, not possible to analyse the pattern ofinterdependency of migration with income, poverty, employmentstatus, education and locational characteristics based on this dataas the census does not collect information on most of the economiccharacteristics. Also, the relevant information is not available atthe household or individual level. Furthermore, a significantproblem has also been noted in case of the migration data fromPopulation Census of 2001. The growth rate of migrants duringthe 1990s with less than a 10-year duration has been observedto be significantly below that with 10 to 20 years and more than20 years of duration. This clearly reflects an erroneous reportingof the duration of stay at the place of enumeration by recentmigrants, understandably due to certain benefits relating to tenurialstatus, access to amenities, etc, being linked to the date of arrival,particularly in case of urban areas. In view of these, the presentstudy restricts itself primarily to the analysis of the data fromthe 55th round of NSS. One big advantage of using this dataset is that, for the first time, a migration schedule was canvassedalong with consumption expenditure and employment in thisround. Understandably, a similar analysis cannot be carried outwith the more recent quinquennial data from the 61st round, asthe information on migration is not tagged on to it. The papernonetheless uses figures from previous rounds for assessing thetrends and patterns at the macro level.The analysis begins with simple cross classification of thenumber of migrants and households as available from the pub-lished reports of NSS, by their socio-economic, including locationalcharacteristics. It is, however, evident that the published data arehighly aggregative and do not permit an in-depth analysis of theinterdependencies of these characteristics with mobility behaviour.The tables have therefore been generated for all persons in urbanareas, using the unit level data from the 55th round of NSS,focusing on poverty and mobility factors and their relationshipwith consumption expenditure, size of household, age, educa-tional levels and employment attributes, including the numberof subsidiary activities. Furthermore, a logistic regression modelhas been applied by taking all individuals in the age group from15 to 65 years who are in the labour force, as observational units.Through this, an attempt has been made to estimate the probabilityof being poor as a function of migration status, level of education,city size of his/her residence, etc, independently, after controllingfor the influence of other factors. For example, the model attemptsto ascertain the impact of city size on the incidence of povertyby controlling for the effect of education and migration. In casethe location of a household/individual is noted to have an effecton the probability of being poor even under this control situation,one can argue that large cities exert an influence on the incidenceof poverty over and above education and migration status of thepopulation [Dubey, Gangopadhyay and Wadhwa 2001]. A simi-lar analysis has been carried out in case of other factors as well.It may be argued that household size captures the dependencyfactor and consequently, larger household size would implyTable 1: Percentage Poor in Different Size Classesof Cities/TownsCity/Town Size1987-881993-941993-941999-2000Large towns/cities35.222.618.414.2Medium cities/towns40.532.227.620.4Small towns45.336.233.224.2All urban centres41.231.427.419.9Rural areas47.641.035.723.9Source:Kundu and Sarangi (2005).
Economic and Political WeeklyJanuary 27, 2007301greater probability of falling into poverty. Conversely, with anincrease in age, a person would shift to higher income/consump-tion categories and, hence the probability of her falling belowthe poverty line would decline, given the employment status andother social characteristics.Migration status has been included in the explanatory modelto ascertain whether the in-migrants or the seasonal migrants havea greater probability of falling below the poverty line and if thatvaries depending on the source (rural or urban) of migration. Ingeneral, it is hypothesised that migration of a person to urbanareas for economic reasons expands her earning opportunities,and hence must have a negative impact on poverty. One mayfurther stipulate that the relatively better off among the ruralpopulation are able to move to cities and towns, since in orderto get a foothold in urban areas, one requires initial investment,certain skill levels and capacity to handle new technology. Also,the capacity to migrate places the person at a higher pedestalthan the locals (if it is a free and conscious decision) and helpsin increasing her earnings. However, if urban poverty is a spilloverof rural poverty [Dandekar and Rath 1971], one would expecta higher incidence of poverty among the migrants than the non-migrants, since the average urban income happens to be signifi-cantly above that in rural areas.The level of education is stipulated to have a negative impacton poverty since it enhances income earning potential, derivedthrough increased access to factor markets and also higher factorproductivity. Similarly, the number of subsidiary activitiesundertaken by a person diversifies the income stream and mayreduce the risk of falling into poverty. It may, however, be arguedthat a person engages himself in subsidiary activities when herprincipal source of earnings is not sufficient for satisfactory livingconditions. Therefore, it would be interesting to see the net effectof the two factors pulling in opposite directions.A brief description of the variables in the model is in order.The dependent variable “poverty” is a binary variable and it takesa value “1” for persons belonging to households falling belowthe poverty line and “0” otherwise. With the threefold classi-fication of urban centres, as presented above, two dummies havebeen introduced for capturing the effect of city size. “LargeD”takes the value “1” if a person resides in large cities and “0”otherwise. The dummy for medium cities/towns denoted as‘MediumD’ follows the same logic. One would expect that theregression coefficients associated with both the dummies to benegative and significant as per the hypothesis that with increasein city size the probability of being poor goes down. A discreteindicator, level of education “edulev” has been incorporated,which varies between 1 and 5, the lowest value 1 assigned toilliterates and the highest value 5 referring to graduates and above.Two indicators on in-migration (excluding the persons fromoutside the country) are introduced on the basis of place of lastresidence of the migrant, i e, rural to urban migrants (RU_mig)and urban to urban migrants (UU_mig). Both the indicators takethe value of 1 and 0 for migrants and non-migrants respectively.Similarly, “S_mig” is a binary variable taking the value of unityfor a seasonal migrant and zero otherwise. Further, to capturethe differential impact of the above mentioned variables acrosscity size, the model incorporates interactive dummies of thesevariables (T1edulev, T2edulev, T1ru_mig, T2uu_mig) along withthe city size dummies. If an explanatory variable in the modelhas a differential impact on the dependent variable (probabilityof being poor) depending on the city size of the person, one wouldexpect the coefficient of these dummies to be significantly differentfrom zero. The results of logit estimations for different groupsof individuals and the aggregative sample are presented in Table7.The sample individuals selected out of the NSS data set forregression analysis are all in the labour force and consequentlybelong to any one of the four categories, namely (a) self-employed,(b) regular wage workers and salaried persons, (c) casual labourers,and (d) unemployed persons. Five different sets of regressionshave been run by taking the individuals belonging to the abovefour categories, as also the aggregative sample of all the fourgroups, as observational units. The persons who could not beplaced in any of the four categories have been excluded fromthe regression analysis.A major limitation confronting this exercise is the samplingdesign of NSS, which is supposed to be appropriate for generatingestimates of consumption expenditure and poverty only at thestate and (NSS) region levels. Recent publications of NSS pointout that as a result of inadequate sample size (largely due todifficulties in increasing the field staff), the estimates have hadhigh standard errors and consequently low reliability, in a largenumber of states. It is difficult to overcome this limitation unlessthe sample size is increased. Without that, the identification ofthe factors explaining the incidence of poverty for different sizeclass of urban centres at the state level would have problems ofreliability. These would, however, be less vulnerable to samplesize and report lower standard error if obtained only at the nationallevel. Keeping this in view, the present paper analyses the variationsin the incidence of poverty and for different size class of townsonly at the national level.IIIPatterns of MigrationA cross classification of migration data across consumptionexpenditure categories reveals that at the macro level, economicdeprivation is less of a factor in migration of men (migrationof women being determined largely by socio-cultural factors),both in rural and urban areas. The migration rate is as high as23.3 per cent in the category with the highest monthly per capitaexpenditure (MPCE), which goes down systematically, the ratebeing as low as 4.3 in the lowest class in rural areas (Table 2).The same is valid in urban areas as well, the correspondingpercentage figures being 43.3 and 10.5. This proposition hasTable 2: Migration Rate for Rural and Urban Males in DifferentMPCE Classes, 1999-2000RuralUrbanMPCE ClassesPer CentMPCE ClassesPer Cent(Rs)Migrants(Rs) Migrants0-2254.30-30010.5225-2553.7300-35013.0255-3004.0350-42513.4300-3404.6425-50019.7340-3804.9500-57521.1380-4205.8575-66523.9420-4706.3665-77527.8470-5257.3775-91530.7525-6158.6915-112037.1615-77510.71120-150041.2775-95014.51500-192538.8950 and above23.31925 and above43.3All6.9All25.7Source: NSS Report No 470: Migration in India, 1999-2000.
Economic and Political WeeklyJanuary 27, 2007302further been validated for the urban areas by tabulating the unitlevel data across five quintiles for the total population, as presentedin Table 3. It reveals that the share of the highest quintile in thetotal number of immigrants is over 26 per cent – much aboveits 20 per cent share in population. Correspondingly, the shareof each of the bottom two quintiles is significantly below 20 percent, reflecting lesser mobility in lower expenditure categories.The persons in urban areas who have gone to any other placefor 60 days or more during the last six months from the dateof survey and returned back have been termed as seasonal orshort duration migrants for the purposes of our analysis. A largesegment of them are possibly those who adopt coping strategiesfor livelihood and survival by shifting to other places in leanseasons. One would then stipulate a positive association of seasonalmigrants with poverty. This short duration movement, on theother hand, can be due to factors like periodic transfer of regularworkers, temporary posting of marketing and extension workers,etc. Interestingly, the migration pattern in Table 3 reveals thatpoverty is not the key factor behind seasonal migration. Indeed,this mobility is not very high among the poor when comparedto middle class households. The bottom 40 per cent of thecountry’s urban population account for only 29 per cent of thetotal seasonal migrants. In contrast, the share of the third quintileis 29 per cent, much above its population share. All these suggestthat even such short-term migration opportunities in urban areasare being cornered by the well-off sections.Poverty among urban households classified by the number ofmembers reporting mobility brings out yet another dimensionof social dynamics (Table 4). It is evident that the pooresthouseholds are those that have one or a few of their membersreporting immigration status. However, when all the membersare in-migrants, the households are observed to belong to eco-nomically better-off strata. These households are in fact moreaffluent than the non-migrant households as the incidence ofpoverty here is the lowest.Understandably, poverty among the salaried persons or thosein regular employment is the lowest (Table 5). The next lowestfigure has been reported, not very surprisingly, by the unemployedpersons. This is a reflection of the capability of these personsto stay out of the labour market (linked to their assets, savings,etc) as they can afford to wait for appropriate jobs. Casual workersreport a very high level of poverty that should be a matter ofconcern for the architects of the National Rural EmploymentGuarantee Scheme, which excludes the urban areas. The highestpoverty figure, however, is recorded by the persons classifiedas others, comprising those outside the labour force. As largesections of these people are children and aged dependents, theirpoverty figures are expectedly very high. Indeed, the householdsthat have a large number of dependents have a greater risk offalling below the poverty line.Table 6 exhibits a negative relationship of the incidence ofpoverty with levels of education, as the former declines smoothlyas one moves from illiteracy to graduation and above. This couldTable 3: Distribution of Urban Migrants across Quintiles ofConsumption Expenditure in 1999-2000Quintile GroupIn-MigrantsSeasonal MigrantsLowest14.213.92nd17.515.13rd19.229.04th22.919.1Highest26.122.8Total100100Source:Calculated from unit record data pertaining to the 55th rounds ofEmployment and Unemployment Surveyof NSS.Table 4: Households Classified by All or a Few MigrantMembers and Incidence of Poverty SamplePer CentHouseholdsHouseholds(Per CentBelow Povertyto Total) LineNon-migrants30.714.0Household having migrant members50.419.3All household members being migrants18.95.6Total10015.1Source:Same as Table 3.Table 5: Proportion of Population Below Poverty Line byEmployment Status (All Persons) Share of PopulationPer Cent PopulationBelow Poverty LineSelf-employed14.219.3Regular13.58.7Casual6.137.9Unemployed1.715.5Others64.620.9Total10019.9Source:Same as Table 3.Table 6: Proportion of Population Below Poverty Line by Levelof Education (All Persons)Education levelPer cent Share inPer Cent PopulationPopulation Below Poverty LineIlliterate27.935.4Up to primary30.322.2Up to secondary26.011.0Up to higher secondary6.95.4Graduate and above8.92.1Total10019.9Source:Same as Table 3.Table 7: Percentage of Migrants to Total Population in Different Size Class of Urban CentresClassified by Their Levels of Education in 1999-2000IlliteratesUp to PrimaryUp to SecondaryUp to Higher SecondaryGraduate and aboveTotalIn-migrants Large towns/cities32.225.732. cities/towns33.927.439.443.849.935.5Small towns31.026.536.741.449.132.6All urban areas32.526.736.639.744.033.3Seasonal-migrants Large towns/cities1. cities/towns0. towns1. urban areas1. as Table 3.
Economic and Political WeeklyJanuary 27, 2007303be a manifestation of the economic pay-off of education in theurban labour market in a period of globalisation. This could alsobe due to the capability of richer sections of population to sendtheir children to schools and higher institutions of learning.The percentage of migrants, both total and seasonal, withdifferent levels of education has been computed in different sizeclass of urban centres by taking the corresponding populationfigure in the denominator (Table 7). One would argue that thereis no variation in the incidence of seasonal migration acrosseducational categories when all urban centres are taken together.This, however, is not the case as far as the total in-migrants areconcerned. The percentage figure goes up from 32 per cent forilliterates to 44 per cent for graduates, in case of all urban centres.The inclination of people with higher education towards migra-tion emerges much more sharply in case of medium and smalltowns as these report about 50 per cent of their graduates to bein-migrants, as opposed to the figure of 34 per cent for millionplus cities.3 The same can be noted as true for seasonal migrantsas well, since the figure for graduates (though small) is notedas almost twice as high as in the case of illiterates. It is onlyin large cities that the percentage of seasonal migrants amongilliterates is significantly above that of the graduates.IVEmerging Interdependenciesand Their ImplicationsWe begin the discussion here by looking at the results pertainingto the total sample, comprising the three categories of employedand that of unemployed. The coefficients of the city size dummiesemerge as negative and significant at 1 per cent level, even aftercontrolling for all other explanatory variables (Table 8). It impliesthat the incidence of a person being poor is less in large citiescompared to medium and small towns, irrespective of themigration status, age, number of subsidiary activities undertaken,etc. This is true for persons with different levels of educationincluding the illiterates. Further, the coefficient of the dummypertaining to large cities is higher than the one for medium cities.One could infer that shifting to medium size cities reduces theprobability of falling into poverty by a factor 0.486, while incase of large cities, it is reduced by a factor 0.782. This confirmsthe hypothesis that the probability of a person being poor is lowif the person is in a metro city compared to any other urban centre.Analysing for different categories of employed separately, itis noted that the pattern of probability for self-employed is similarto that of the total sample. For regular wage workers, however,the city size effect is insignificant. This is because most of thehouseholds with one salaried person would have a per capitaincome or consumption above the threshold level that definesthe poverty line, in all size class of urban centres. In case of casualworkers, probability of being poor declines significantly in higherorder towns compared to small cities. The large cities, however,do not seem to have an advantage over medium towns in this regard.The pattern, however, is different for unemployed persons; thelarge cities provide diverse opportunities of livelihood, therebyreducing the incidence of poverty even among those “enumeratedas unemployed”, when compared to medium and small towns.Age of a person has a negative effect on the incidence of povertyin the aggregative sample and also for the self-employed andregular workers. The effect, however, does not seem to besignificant in case of casual workers and unemployed persons.It may be argued that for self-employed and regular workers,age helps in settling down in jobs and acquiring skills that increasetheir earnings and reduce the probability of falling into poverty.This, unfortunately, is not the case of casual workers as theirwages remain subject to market fluctuations, the age having nopositive impact on their earning. Age of a casual worker is thusnot an important factor in poverty reduction.Table 8: Determinants of Poverty in the Year 1999-2000: Logit Regression ResultsAll PersonsSelf-EmployedRegular/SalariedCasualUnemployed(Sample=78566) (Sample=32702) (Sample=29781) (Sample=12050) (Sample=4033)CoeffP-valueCoeffP-valueCoeffP-valueCoeffP-valueCoeffP-valueLarge D-0.7820.00-1.0810.00-0.2800.35-0.0670.87-1.5440.03Medium D-0.4860.00-0.6210.00-0.2760.14-0.3220.03-0.6510.14Age-0.0140.00-0.0130.00-0.0150.00-0.0010.680.0010.92Edulev-0.8550.00-0.8620.00-0.7390.00-0.5750.00-0.5570.00T1edulev-0.0190.800.0610.51-0.1970.04-0.2710.070.3210.30T2edulev0.0490.110.1190.01-0.0230.720.0120.870.0320.81Members0.2230.000.1980.000.2620.000.2980.000.3020.00Subact0.1560.000.0540.380.3310.040.2060.01 ––RU_mig-0.1970.00-0.1310.11-0.3730.01-0.1100.28-0.1450.75T1ru_mig-0.2870.04-0.0100.97-0.2300.39-0.5010.06-0.6510.47T2ru_mig-0.0370.64-0.0990.400.0040.99-0.0160.910.0300.96UU_mig-0.8570.00-0.5730.00-1.1630.00-0.7960.00-0.2510.55T1uu_mig0.2780.240.5480.130.4770.33-0.2520.52-2.3230.06T2uu_mig0.2330.050.0510.780.4910.070.2470.32-0.1010.85S_mig0.0550.710.0090.97-0.2880.290.3400.25-0.0420.94_cons0.1350.100.0990.39-0.6910.00-0.7170.00-1.5580.00Notes:Dependent variable: Poor, takes the value 1 if the household is poor, 0 otherwiseLarge D –Dummy for large citiesMedium D –Dummy for medium citiesAge – Age of the person (the sample varies between age 15 years and 65 years)Edulev –Level of education of principal earning member of householdSubact –No of subsidiary activities undertaken by the persons in the householdMembers–Household sizeS_mig – Seasonal (short-duration) migrant: a person who has gone out for work for 60 days or more during the last 6 monthsRU_mig – Migrating from Rural (by place of last residence) to UrbanUU_mig – Migrating from Urban (by place of last residence) to UrbanT1edulev and T2edulev – Interaction term of of city size dummies for large and medium towns respectively and level of educationT1ru_mig and T2ru_mig – Interaction term of city size dummies for large and medium towns with migration from rural to urban (RU_mig)T1uu_mig and T2uu_mig – Interaction term of city size dummies for large and medium towns with migration from urban to urban (UU_mig)
Economic and Political WeeklyJanuary 27, 2007304The level of educational attainment seems to be the single mostsignificant factor impacting on poverty, for all and also for thefour different categories of labour force, mentioned above. Atthe aggregative level, one would note that with increase in levelof education, probability of being poor reduces by a factor 0.855.Understandably, the decline is relatively less in case of unem-ployed persons and casual labourers, the values being 0.557 and0.575 respectively.The interactive dummies of large and medium cities witheducation do not show any differential impact at the aggregativelevel. Looking at the results across employment categories, onewould notice that the coefficient of these dummies for regular/salaried and casual workers are negative and significant in largecities, while they are not significant in medium cities. One wouldinfer that persons with a high level of education in large citiesare likely to have higher earnings and the combined effect ofcity size and education would enhance the chances of escapingfrom poverty. Unfortunately, that is not the case in medium andsmall towns. Moreover, an educated self-employed person seemsto be more vulnerable in medium cities, the regression coefficient(of the interaction dummy) being positive and significant, whiletheresult is not significant in large cities. One would, therefore, inferthatan educated person in the medium cities would have difficultiesin improving her economic status through self-employment.The size of household has a positive impact on the risk of itsmembers falling into poverty, the coefficient being 0.223 for allpersons. This means, larger the size of the household, higher theprobability of its being poor. The pattern remains similar acrossall employment categories. Similarly, the number of subsidiaryactivities undertaken by an individual has a strong and positiveeffect. Here again, the pattern is identical across all employmentcategories, excepting self-employment. It may therefore be arguedthat a person with one or more subsidiary activities is morevulnerable to poverty. Understandably, it is the weak socio-economic condition of the household that forces its members toseek multiple employment. Indeed, a household goes for diver-sification of activities only when it is not getting adequateearnings from its principal source.The coefficients of rural to urban migration (RU_mig) beingstrong and negative for the aggregate sample, as also for regular/salaried persons, support the proposition that RU mobility is afactor in poverty reduction. The coefficients are generally nega-tive for other labour force categories as well, although all of theseare not statistically significant. This puts a question mark to theearlier hypothesis that urban poverty is the spillover of ruralpoverty,4 since the rural migrants into urban areas have alower probability of being poor than the local population. It ishowever possible that migration itself is the factor responsiblefor increasing the earnings of individuals, enabling them to goover the poverty line. Alternately, one may argue that it is largelythe relatively better off sections who are able to migrate to urbancentres since moving to cities requires initial staying capacityand certain levels of skill and, consequently, poverty is low amongRU migrants. Indeed, with modernisation and technologyupgradation in many of the urban sectors, the absorption of therural poor has become increasingly difficult. The likelihood offalling into poverty is low in case of UU migrants as well. Thepattern is similar across all employment categories, except forthe unemployed, which is understandable. Migration, both fromrural and urban areas, thus, emerges clearly as an instrument ofimproving economic well-being and escaping poverty for theadult population.Noticeably, seasonal migrantion does not show any significantimpact in terms of increasing or reducing the probability of fallingbelow poverty line for any of the four categories considered inthe analysis. This supports the hypothesis, noted in the precedingsection that seasonal (short duration) migration is not restricted tothe rural poor struggling for survival, but is also common amongbetter off households, shifting temporarily for better opportuni-ties or posting to another village or town for a short duration.Households reporting seasonal migration are, thus, socially andeconomically heterogeneous and consequently, no pattern acrossexpenditure categories has been reported as significant.The probabilities for individuals, with different employmentstatus, to fall below the poverty line have been calculated fromthe logit regression model (excluding the seasonal migrants andthe interactive dummies), at the sample means of differentexplanatory variables (Table 9). All the coefficients in the modelwork out as significant with the constant term (intercept) reportedas insignificant. The estimated probabilities that have been reportedfor only two of the educational categories in Table 9 are thusamenable to socio-economic interpretation.One would note that the probabilities of being poor for therural to urban (RU) migrants, who are illiterates, are very high,much above that for higher levels of education. For the graduatesamong them, however, the probability is strikingly low. Amongthe illiterates, non-migrants have a greater poverty risk comparedto RU migrants, the smallest values being reported by UU migrants.Furthermore, regular workers among illiterate migrants reporta lower chance of being below the poverty line compared to theself-employed. The highest probabilities are observed in the caseof casual workers in all size class of towns and migrant or non-migrant categories. Also, the pattern confirms the general patternthat the probabilities are the least in large cities, followed bymedium and then by small towns. This pattern is found to bevalid across all employment categories.Table 9: Estimated Poverty Probabilities across Different Type of Household in Different Size Class of Cities/Towns Self-empRegularCasualUnempTotalSelf-empRegularCasualUnempTotalIlliterate RU MigrantsGraduates+ RU MigrantsLarge0.2040.1730.3660.2490.2140.0130.0100.0280.0170.014Medium0.2860.2470.4750.3420.2990.0200.0160.0280.0260.021Small0.3620.3160.5610.4240.3760.0280.0230.0610.0360.030 Illiterate UU MigrantsGraduates+ UU MigrantsLarge0.1480.1240.2820.1840.1560.0090.0070.0200.0110.009Medium0.2140.1820.3810.2610.2250.0140.0110.0300.0180.014Small0.2780.2400.4650.3330.2910.0190.0160.0420.0250.020 Illiterate Non-migrantsGraduates+ Non MigrantsLarge0.2460.2110.4240.2980.2580.0160.0130.0360.0210.017Medium0.3390.2950.5360.3990.3530.0250.0210.0550.0330.027Small0.4200.3710.6200.4840.4350.0350.0290.0760.0450.037
1 2 3 4 5 Education RU Migrants Non-Migrants UU Migrants

Dear Reader,

To continue reading, become a subscriber.

Explore our attractive subscription offers.

Click here

Back to Top